Center-specific Federated Learning for Radiation Pneumonitis:A Cross-Center Adaptive Alternating Framework.
Authors
Affiliations (11)
Affiliations (11)
- Department of Radiation Oncology, Key Laboratory of Cancer Prevention and Therapy, Tianjin Medical University Cancer Institute & Hospital, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Tianjin, 300060, China; Department of Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands.
- Department of Ultrasound, Beijing Friendship Hospital, Capital Medical University, Beijing 100050, China; Precision and Intelligence Medical Imaging Lab, Beijing Clinical Research Institute, Beijing Friendship Hospital, Capital Medical University, Beijing, 100050, China.
- Department of Radiation Oncology, Shanghai Chest Hospital, School of medicine, Shanghai Jiao Tong University, Shanghai,200030.
- Department of Radiation Oncology, Zhejiang Cancer Hospital; Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences; Zhejiang Key Laboratory of Particle Radiotherapy Equipment, Hangzhou, China.
- Department of Radiation Oncology, Key Laboratory of Cancer Prevention and Therapy, Tianjin Medical University Cancer Institute & Hospital, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Tianjin, 300060, China.
- Department of Radiation Oncology, The Second People Hospital of Dezhou, No.55 Fangzhistreet, Dezhou City, Shandong Province, 253000, China.
- Shenzhen University Medical School, Shenzhen University, 3688 Nanhai Avenue, Shenzhen, 518060, P. R. China.
- Data Science Institute (DSI), Hasselt University, Hasselt, Belgium.
- Department of Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands.
- Department of Radiation Oncology, Key Laboratory of Cancer Prevention and Therapy, Tianjin Medical University Cancer Institute & Hospital, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Tianjin, 300060, China. Electronic address: [email protected].
- Department of Radiation Oncology, Zhejiang Cancer Hospital; Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences; Zhejiang Key Laboratory of Particle Radiotherapy Equipment, Hangzhou, China; Department of Radiation Oncology, Key Laboratory of Cancer Prevention and Therapy, Tianjin Medical University Cancer Institute & Hospital, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Tianjin, 300060, China; Department of Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands. Electronic address: [email protected].
Abstract
Accurate prediction of symptomatic radiation pneumonitis (RP) is critical for radiotherapy, yet the generalization of deep learning models is hindered by restricted access to multi-center data. Although federated learning (FL) bypasses data-sharing restrictions, standard FL algorithms underperform on highly heterogeneous clinical data across institutions. Therefore, this study aims to evaluate the clinical feasibility of a center-specific federated learning approach. We evaluated the Federated Cross-Center Adaptive Alternating Model (FCAAM), a tailored framework designed to decouple globally transferable representations from the center-specific adaptations. The framework uses a dynamic weighting mechanism to handle data heterogeneity and employs differential privacy for enhanced security. The proposed FCAAM was evaluated for the prediction of RP using planning CT and dose images on a diverse cohort of 1,238 patients from four datasets representing real-world temporal and spatial shifts. Its performance was compared with single-center model, centralized model, and standard federated average model (FedAvg). FCAAM demonstrated improved cross-center performance and consistent robustness compared to baseline. It achieved a stable area under the curve (AUC) across all four dataset test sets (0.71-0.77), outperforming the single-center models (all AUCs < 0.70) and FedAvg. FCAAM's performance was comparable to the centralized model and showed a relative improvement in sensitivity to small-sized datasets. Interpretability analysis confirmed that FCAAM learned clinically relevant features, and a web platform demonstrated the practical feasibility of applying FCAAM for multi-center collaboration. FCAAM provides a privacy-preserving, robust and interpretable solution for multi-center RP prediction. This center-specific strategy shows potential to enhance clinical decision-making and reduce cross-center performance gaps, supporting safer personalized radiotherapy.